The universal design of roof photovoltaic (PV) systems in substations often neglects specific roof resources and regional irradiance characteristics, which frequently leads to capacity waste. To address this issue, this paper focuses on 220 kV fully indoor substations to study the optimal capacity configuration of roof PV systems. Based on a survey of PV construction data across multiple provinces, the characteristics of roof resources and the current status of PV configurations are analyzed. Furthermore, the principles for module selection, array design, and layout are clarified. A PV output power model incorporating inclined-plane irradiance and module operating temperature is constructed, and typical PV output scenarios are generated using the K-medoids clustering algorithm. With the objective of maximizing comprehensive economic benefits over the entire life cycle, an optimal configuration model is established, integrating constraints on spatial limits, power balance, and anti-backfeeding. Finally, a 220 kV fully indoor substation in Jiangsu is taken as a case study to determine the optimal configured capacity. The proposed method provides a valuable reference for the PV capacity configuration of different substations, effectively improving design rationality and economic benefits.